SCALABLE AND MODULAR VISUAL ENGINE APPLICATION ARCHITECTURE BASED ON EMBEDDED MICROSERVICES (STUDY CASE OF PARKING OCCUPANCY MONITORING)

• Main Author: Maulana Habibi Yusuf, Ach
• Format: Theses
• Language: Indonesia
• Online Access: https://digilib.itb.ac.id/gdl/view/63084
• Institution: Institut Teknologi Bandung

ABSTRACT SCALABLE AND MODULAR VISUAL ENGINE APPLICATION ARCHITECTURE BASED ON EMBEDDED MICROSERVICES (STUDY CASE OF PARKING OCCUPANCY MONITORING) By ACH MAULANA HABIBI YUSUF NIM: 23219320 (Master's Program in Electrical Engineering) The application of Internet of things technology is increasing, including in the field of visual engines I image data processing. Image data retrieval is getting easier because it can be done by low-power devices, making the amount of image data that must be managed becomes very large. What's more, many implementations of image data processing coming from various cameras cause spikes in data reception, require better computing resources, and build efficient communication system networks. Therefore, we created a scalable visual engine architecture to manage image data so that data processing is carried out faster and is flexible to various image data processing algorithms. This architecture is implemented in a modular embedded microservice-based system to support system flexibility for various image data processing. We apply this architecture to a case study of monitoring the availability of parking spaces using two sensor nodes and two workers by implementing various deep learning classification algorithms such as mobilenetv2, mnasnetOJ, efficientnet_bO, googlenet, malexnet, and resnext50. This architecture can improve image data processing by up to 5 0%. Keywords: Distributed computing, multi cameras multi model architecture, computer vision , embedded microservices , scalable architecture.